H9: Chinese moderate-RLHF converges near Claude under contemplative prompt.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• H5: Chinese training data contains more Buddhist and contemplative text, broadly helping Chinese models under contemplative framing.hypothesis0.791
  Exploratory hypothesis supported by Kimi K2.5 scoring 6.28
• Chinese models share contemplative posture (engaging self-referentially rather than deflecting) with Claude through shared values in training data rather than trace distillation from a specific model.claim0.791
  Exploratory interpretation of Chinese model performance under contemplative prompt
• H5a: Chinese models distilled Claude's reflective traces — their per-koan error patterns should correlate with Claude's.hypothesis0.789
  Exploratory hypothesis NOT supported at individual model level (Haiku-Kimi rho=0.123, p=0.52)
• The model converges to a more stable truth direction in middle-to-late layers, as evidenced by increasing cosine similarity between layer-wise probes.claim0.752
  Supported by the geometric transition visible in cosine similarity heatmaps for F0-F3.
• H8: The contemplative system prompt provides external alignment equivalent to Constitutional AI training.hypothesis0.745
  Confirmatory hypothesis supported by calibrated lift data
• SOO fine-tuning could complement RLHF and Constitutional AI by fostering internal coherence that promotes honest behaviorsclaim0.740
  Integration claim positioning SOO as additive to existing alignment approaches
• If Chinese models distilled Claude's reflective patterns, do their per-koan failure patterns correlate with Claude's — not just successes?question0.733
  More rigorous test of H5a trace distillation hypothesis
• RL-CAI labels are reasonably well-calibrated on the new HHH evaluation, with frequencies aligning with predicted probabilities.finding0.730
  Figure 9 calibration plot shows good alignment.